General > General Technical Chat
Looking for: Manageable Gbit ETH switch for 10" rack (8port+SFP, with PoE)
Yansi:
--- Quote from: tszaboo on December 31, 2021, 02:05:46 pm ---
--- Quote from: Ranayna on December 23, 2021, 12:39:03 pm ---Indeed it is.
Though i think what tszaboo means is, that virtually no switch is able to provide full power to every port that it has. They all have a power budget cap significantly lower than the theoretical maximum.
But that is fine in most cases. It is relatively rare for IT equipment to really draw full power. For example our IP phones require around 6 to 7 watts in standby, with the display turned on, and not really much more while in a call. A lot lower than the ~15 Watts available with 802.3af. Similar with our WiFi 6 Access Points. Those stay below 20 watts, again lower than the ~30 watts of 802.3at.
At least that are the values that our switches show me on the cli, and the switch also uses these actual values for the calculation of power budgets.
--- End quote ---
True. PoE standard can be anywhere from 0W to ~75W as I recall per port. So a 16 port switch would need to have 1KW power supply built in. Plus the magnetics is more expensive for PoE because it can go into DC saturation. And normal PoE is a smart protocol, you need an adjustable DC-DC per port. All of these cost money.
--- End quote ---
Nobody forces you to support 75W a port. Heck nobody even forces you to be able to supply all ports at once at full power. Most of the cheaper PoE switches support 802.3at on all ports, but will not deliver that much power on all ports at once, as the power budget is quite smaller.
And as far as I know you do not need any adjustable DC/DC per port. At least not for 802.3af or 802.3at I am a bit familiar with. On the source side, there is just a mosfet to switch it on or off and some other smarts (metering, protection and device identification), that can fit easily multiple channels within a single small TSSOP/QFN IC.
The (isolated) DC/DC is on the powered device side, where you need to provide the galvanic isolation. I do not think that many (if any) PoE switches provide galvanically isolated DC output in between ports (power source side). The power is running simply -48 to some -53V against a common ground - so the ports on the source equipment are not galvanically isolated in between each other, they run with a common ground of the PoE supply.
Please correct me if I am wrong, I have only studied the 802.3af/at only so far, never done any actual designs with it yet so I may have missed some requirements.
mansaxel:
The Lancom says "Orchestrated from the LANCOM Management Cloud, it is configured efficiently and automatically by SD-LAN." which means that it never truly is yours. You never know if the switch is going to be a brick tomorrow from planned obsolescence because Vendor wants to "refocus" or wants you to buy a new model.
Since you already HAVE a 19" rack, buy a 19" switch and be done with it.
Fully agree with the notion that this is something that should be easy to swap out, because in 3 years the world will have moved on, and placing the restraint 'must be 10"' on a COTS item is going to be cumbersome, as you've already discovered.
ve7xen:
--- Quote from: Ranayna on December 23, 2021, 12:39:03 pm ---Indeed it is.
Though i think what tszaboo means is, that virtually no switch is able to provide full power to every port that it has. They all have a power budget cap significantly lower than the theoretical maximum.
But that is fine in most cases. It is relatively rare for IT equipment to really draw full power. For example our IP phones require around 6 to 7 watts in standby, with the display turned on, and not really much more while in a call. A lot lower than the ~15 Watts available with 802.3af. Similar with our WiFi 6 Access Points. Those stay below 20 watts, again lower than the ~30 watts of 802.3at.
At least that are the values that our switches show me on the cli, and the switch also uses these actual values for the calculation of power budgets.
--- End quote ---
Just be careful, because the allocated power is (should) be calculated based on the advertised power requirement of the PD, not its actual consumption, for obvious reasons (it needs to avoid being overloaded if usage increases, which is a bad situation that would require cutting power from an active PD). It's pretty easy to allocate more than the available budget while consumption remains considerably lower, and the switch will (should) reject additional allocation requests from PDs in this situation.
--- Quote ---Looks like a toy, but is no toy. Unfortunately I do not have use for 2.5Gb or 10Gb SFP, as most of the connected instrumentation will be 100M. I know, good to future proof it, but I am pretty sure I do not need this here. Also not Cheap.
--- End quote ---
You may not need it, but many active/non-obsolete devices in this segment are going to have 10G uplinks, it's the modern expectation. Such devices may still suit your purposes.
I think mikrotik CRS112-8P-4S-IN (with the optional 48V power supply) may suit your requirements.
Yansi:
Yes, sure, I have a 19" rack. But that has some limited number of Us, where I need to fit my stuff.
And if manufacturers were not morons and would conform to the 10" standard, no problems like this would exist. There is a metric sh!t ton of "less than 19 yet rack mountable" devices, but none conform to the 10" standard, even though the boxes are almost full of air and could be easily designed to fit within 10" enclosure.
Yes, I am crying on a wrong grave here... but... I just like to share my sheer disappointment with finding a 10" switch to be that difficult.
And BTW, I have already bought the TPlink TL-SG2210P, and already received it. And I am very satisfied. So kudos to rob77 for sending a tip for this one. Two more pieces already on the way home.
The build quality inside seems very decent. PCB layout looks very nice. Attaching few photos to please an eye or two. The web interface is very nicely layed out, easy to navigate and responsive.
Navigation
[0] Message Index
[*] Previous page
Go to full version